[NETFILTER]: ctnetlink: use netlink policy
[linux-2.6.git] / net / ipv6 / netfilter / nf_conntrack_l3proto_ipv6.c
1 /*
2  * Copyright (C)2004 USAGI/WIDE Project
3  *
4  * This program is free software; you can redistribute it and/or modify
5  * it under the terms of the GNU General Public License version 2 as
6  * published by the Free Software Foundation.
7  *
8  * Author:
9  *      Yasuyuki Kozakai @USAGI <yasuyuki.kozakai@toshiba.co.jp>
10  */
11
12 #include <linux/types.h>
13 #include <linux/ipv6.h>
14 #include <linux/in6.h>
15 #include <linux/netfilter.h>
16 #include <linux/module.h>
17 #include <linux/skbuff.h>
18 #include <linux/icmp.h>
19 #include <linux/sysctl.h>
20 #include <net/ipv6.h>
21
22 #include <linux/netfilter_ipv6.h>
23 #include <net/netfilter/nf_conntrack.h>
24 #include <net/netfilter/nf_conntrack_helper.h>
25 #include <net/netfilter/nf_conntrack_l4proto.h>
26 #include <net/netfilter/nf_conntrack_l3proto.h>
27 #include <net/netfilter/nf_conntrack_core.h>
28
29 static int ipv6_pkt_to_tuple(const struct sk_buff *skb, unsigned int nhoff,
30                              struct nf_conntrack_tuple *tuple)
31 {
32         u_int32_t _addrs[8], *ap;
33
34         ap = skb_header_pointer(skb, nhoff + offsetof(struct ipv6hdr, saddr),
35                                 sizeof(_addrs), _addrs);
36         if (ap == NULL)
37                 return 0;
38
39         memcpy(tuple->src.u3.ip6, ap, sizeof(tuple->src.u3.ip6));
40         memcpy(tuple->dst.u3.ip6, ap + 4, sizeof(tuple->dst.u3.ip6));
41
42         return 1;
43 }
44
45 static int ipv6_invert_tuple(struct nf_conntrack_tuple *tuple,
46                              const struct nf_conntrack_tuple *orig)
47 {
48         memcpy(tuple->src.u3.ip6, orig->dst.u3.ip6, sizeof(tuple->src.u3.ip6));
49         memcpy(tuple->dst.u3.ip6, orig->src.u3.ip6, sizeof(tuple->dst.u3.ip6));
50
51         return 1;
52 }
53
54 static int ipv6_print_tuple(struct seq_file *s,
55                             const struct nf_conntrack_tuple *tuple)
56 {
57         return seq_printf(s, "src=" NIP6_FMT " dst=" NIP6_FMT " ",
58                           NIP6(*((struct in6_addr *)tuple->src.u3.ip6)),
59                           NIP6(*((struct in6_addr *)tuple->dst.u3.ip6)));
60 }
61
62 static int ipv6_print_conntrack(struct seq_file *s,
63                                 const struct nf_conn *conntrack)
64 {
65         return 0;
66 }
67
68 /*
69  * Based on ipv6_skip_exthdr() in net/ipv6/exthdr.c
70  *
71  * This function parses (probably truncated) exthdr set "hdr"
72  * of length "len". "nexthdrp" initially points to some place,
73  * where type of the first header can be found.
74  *
75  * It skips all well-known exthdrs, and returns pointer to the start
76  * of unparsable area i.e. the first header with unknown type.
77  * if success, *nexthdr is updated by type/protocol of this header.
78  *
79  * NOTES: - it may return pointer pointing beyond end of packet,
80  *          if the last recognized header is truncated in the middle.
81  *        - if packet is truncated, so that all parsed headers are skipped,
82  *          it returns -1.
83  *        - if packet is fragmented, return pointer of the fragment header.
84  *        - ESP is unparsable for now and considered like
85  *          normal payload protocol.
86  *        - Note also special handling of AUTH header. Thanks to IPsec wizards.
87  */
88
89 static int nf_ct_ipv6_skip_exthdr(const struct sk_buff *skb, int start,
90                                   u8 *nexthdrp, int len)
91 {
92         u8 nexthdr = *nexthdrp;
93
94         while (ipv6_ext_hdr(nexthdr)) {
95                 struct ipv6_opt_hdr hdr;
96                 int hdrlen;
97
98                 if (len < (int)sizeof(struct ipv6_opt_hdr))
99                         return -1;
100                 if (nexthdr == NEXTHDR_NONE)
101                         break;
102                 if (nexthdr == NEXTHDR_FRAGMENT)
103                         break;
104                 if (skb_copy_bits(skb, start, &hdr, sizeof(hdr)))
105                         BUG();
106                 if (nexthdr == NEXTHDR_AUTH)
107                         hdrlen = (hdr.hdrlen+2)<<2;
108                 else
109                         hdrlen = ipv6_optlen(&hdr);
110
111                 nexthdr = hdr.nexthdr;
112                 len -= hdrlen;
113                 start += hdrlen;
114         }
115
116         *nexthdrp = nexthdr;
117         return start;
118 }
119
120 static int ipv6_get_l4proto(const struct sk_buff *skb, unsigned int nhoff,
121                             unsigned int *dataoff, u_int8_t *protonum)
122 {
123         unsigned int extoff = nhoff + sizeof(struct ipv6hdr);
124         unsigned char pnum;
125         int protoff;
126
127         if (skb_copy_bits(skb, nhoff + offsetof(struct ipv6hdr, nexthdr),
128                           &pnum, sizeof(pnum)) != 0) {
129                 pr_debug("ip6_conntrack_core: can't get nexthdr\n");
130                 return -NF_ACCEPT;
131         }
132         protoff = nf_ct_ipv6_skip_exthdr(skb, extoff, &pnum, skb->len - extoff);
133         /*
134          * (protoff == skb->len) mean that the packet doesn't have no data
135          * except of IPv6 & ext headers. but it's tracked anyway. - YK
136          */
137         if ((protoff < 0) || (protoff > skb->len)) {
138                 pr_debug("ip6_conntrack_core: can't find proto in pkt\n");
139                 return -NF_ACCEPT;
140         }
141
142         *dataoff = protoff;
143         *protonum = pnum;
144         return NF_ACCEPT;
145 }
146
147 static unsigned int ipv6_confirm(unsigned int hooknum,
148                                  struct sk_buff **pskb,
149                                  const struct net_device *in,
150                                  const struct net_device *out,
151                                  int (*okfn)(struct sk_buff *))
152 {
153         struct nf_conn *ct;
154         struct nf_conn_help *help;
155         struct nf_conntrack_helper *helper;
156         enum ip_conntrack_info ctinfo;
157         unsigned int ret, protoff;
158         unsigned int extoff = (u8 *)(ipv6_hdr(*pskb) + 1) - (*pskb)->data;
159         unsigned char pnum = ipv6_hdr(*pskb)->nexthdr;
160
161
162         /* This is where we call the helper: as the packet goes out. */
163         ct = nf_ct_get(*pskb, &ctinfo);
164         if (!ct || ctinfo == IP_CT_RELATED + IP_CT_IS_REPLY)
165                 goto out;
166
167         help = nfct_help(ct);
168         if (!help)
169                 goto out;
170         /* rcu_read_lock()ed by nf_hook_slow */
171         helper = rcu_dereference(help->helper);
172         if (!helper)
173                 goto out;
174
175         protoff = nf_ct_ipv6_skip_exthdr(*pskb, extoff, &pnum,
176                                          (*pskb)->len - extoff);
177         if (protoff > (*pskb)->len || pnum == NEXTHDR_FRAGMENT) {
178                 pr_debug("proto header not found\n");
179                 return NF_ACCEPT;
180         }
181
182         ret = helper->help(pskb, protoff, ct, ctinfo);
183         if (ret != NF_ACCEPT)
184                 return ret;
185 out:
186         /* We've seen it coming out the other side: confirm it */
187         return nf_conntrack_confirm(pskb);
188 }
189
190 static unsigned int ipv6_defrag(unsigned int hooknum,
191                                 struct sk_buff **pskb,
192                                 const struct net_device *in,
193                                 const struct net_device *out,
194                                 int (*okfn)(struct sk_buff *))
195 {
196         struct sk_buff *reasm;
197
198         /* Previously seen (loopback)?  */
199         if ((*pskb)->nfct)
200                 return NF_ACCEPT;
201
202         reasm = nf_ct_frag6_gather(*pskb);
203
204         /* queued */
205         if (reasm == NULL)
206                 return NF_STOLEN;
207
208         /* error occured or not fragmented */
209         if (reasm == *pskb)
210                 return NF_ACCEPT;
211
212         nf_ct_frag6_output(hooknum, reasm, (struct net_device *)in,
213                            (struct net_device *)out, okfn);
214
215         return NF_STOLEN;
216 }
217
218 static unsigned int ipv6_conntrack_in(unsigned int hooknum,
219                                       struct sk_buff **pskb,
220                                       const struct net_device *in,
221                                       const struct net_device *out,
222                                       int (*okfn)(struct sk_buff *))
223 {
224         struct sk_buff *reasm = (*pskb)->nfct_reasm;
225
226         /* This packet is fragmented and has reassembled packet. */
227         if (reasm) {
228                 /* Reassembled packet isn't parsed yet ? */
229                 if (!reasm->nfct) {
230                         unsigned int ret;
231
232                         ret = nf_conntrack_in(PF_INET6, hooknum, &reasm);
233                         if (ret != NF_ACCEPT)
234                                 return ret;
235                 }
236                 nf_conntrack_get(reasm->nfct);
237                 (*pskb)->nfct = reasm->nfct;
238                 (*pskb)->nfctinfo = reasm->nfctinfo;
239                 return NF_ACCEPT;
240         }
241
242         return nf_conntrack_in(PF_INET6, hooknum, pskb);
243 }
244
245 static unsigned int ipv6_conntrack_local(unsigned int hooknum,
246                                          struct sk_buff **pskb,
247                                          const struct net_device *in,
248                                          const struct net_device *out,
249                                          int (*okfn)(struct sk_buff *))
250 {
251         /* root is playing with raw sockets. */
252         if ((*pskb)->len < sizeof(struct ipv6hdr)) {
253                 if (net_ratelimit())
254                         printk("ipv6_conntrack_local: packet too short\n");
255                 return NF_ACCEPT;
256         }
257         return ipv6_conntrack_in(hooknum, pskb, in, out, okfn);
258 }
259
260 static struct nf_hook_ops ipv6_conntrack_ops[] = {
261         {
262                 .hook           = ipv6_defrag,
263                 .owner          = THIS_MODULE,
264                 .pf             = PF_INET6,
265                 .hooknum        = NF_IP6_PRE_ROUTING,
266                 .priority       = NF_IP6_PRI_CONNTRACK_DEFRAG,
267         },
268         {
269                 .hook           = ipv6_conntrack_in,
270                 .owner          = THIS_MODULE,
271                 .pf             = PF_INET6,
272                 .hooknum        = NF_IP6_PRE_ROUTING,
273                 .priority       = NF_IP6_PRI_CONNTRACK,
274         },
275         {
276                 .hook           = ipv6_conntrack_local,
277                 .owner          = THIS_MODULE,
278                 .pf             = PF_INET6,
279                 .hooknum        = NF_IP6_LOCAL_OUT,
280                 .priority       = NF_IP6_PRI_CONNTRACK,
281         },
282         {
283                 .hook           = ipv6_defrag,
284                 .owner          = THIS_MODULE,
285                 .pf             = PF_INET6,
286                 .hooknum        = NF_IP6_LOCAL_OUT,
287                 .priority       = NF_IP6_PRI_CONNTRACK_DEFRAG,
288         },
289         {
290                 .hook           = ipv6_confirm,
291                 .owner          = THIS_MODULE,
292                 .pf             = PF_INET6,
293                 .hooknum        = NF_IP6_POST_ROUTING,
294                 .priority       = NF_IP6_PRI_LAST,
295         },
296         {
297                 .hook           = ipv6_confirm,
298                 .owner          = THIS_MODULE,
299                 .pf             = PF_INET6,
300                 .hooknum        = NF_IP6_LOCAL_IN,
301                 .priority       = NF_IP6_PRI_LAST-1,
302         },
303 };
304
305 #ifdef CONFIG_SYSCTL
306 static ctl_table nf_ct_ipv6_sysctl_table[] = {
307         {
308                 .ctl_name       = NET_NF_CONNTRACK_FRAG6_TIMEOUT,
309                 .procname       = "nf_conntrack_frag6_timeout",
310                 .data           = &nf_ct_frag6_timeout,
311                 .maxlen         = sizeof(unsigned int),
312                 .mode           = 0644,
313                 .proc_handler   = &proc_dointvec_jiffies,
314         },
315         {
316                 .ctl_name       = NET_NF_CONNTRACK_FRAG6_LOW_THRESH,
317                 .procname       = "nf_conntrack_frag6_low_thresh",
318                 .data           = &nf_ct_frag6_low_thresh,
319                 .maxlen         = sizeof(unsigned int),
320                 .mode           = 0644,
321                 .proc_handler   = &proc_dointvec,
322         },
323         {
324                 .ctl_name       = NET_NF_CONNTRACK_FRAG6_HIGH_THRESH,
325                 .procname       = "nf_conntrack_frag6_high_thresh",
326                 .data           = &nf_ct_frag6_high_thresh,
327                 .maxlen         = sizeof(unsigned int),
328                 .mode           = 0644,
329                 .proc_handler   = &proc_dointvec,
330         },
331         { .ctl_name = 0 }
332 };
333 #endif
334
335 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
336
337 #include <linux/netfilter/nfnetlink.h>
338 #include <linux/netfilter/nfnetlink_conntrack.h>
339
340 static int ipv6_tuple_to_nlattr(struct sk_buff *skb,
341                                 const struct nf_conntrack_tuple *tuple)
342 {
343         NLA_PUT(skb, CTA_IP_V6_SRC, sizeof(u_int32_t) * 4,
344                 &tuple->src.u3.ip6);
345         NLA_PUT(skb, CTA_IP_V6_DST, sizeof(u_int32_t) * 4,
346                 &tuple->dst.u3.ip6);
347         return 0;
348
349 nla_put_failure:
350         return -1;
351 }
352
353 static const struct nla_policy ipv6_nla_policy[CTA_IP_MAX+1] = {
354         [CTA_IP_V6_SRC] = { .len = sizeof(u_int32_t)*4 },
355         [CTA_IP_V6_DST] = { .len = sizeof(u_int32_t)*4 },
356 };
357
358 static int ipv6_nlattr_to_tuple(struct nlattr *tb[],
359                                 struct nf_conntrack_tuple *t)
360 {
361         if (!tb[CTA_IP_V6_SRC] || !tb[CTA_IP_V6_DST])
362                 return -EINVAL;
363
364         memcpy(&t->src.u3.ip6, nla_data(tb[CTA_IP_V6_SRC]),
365                sizeof(u_int32_t) * 4);
366         memcpy(&t->dst.u3.ip6, nla_data(tb[CTA_IP_V6_DST]),
367                sizeof(u_int32_t) * 4);
368
369         return 0;
370 }
371 #endif
372
373 struct nf_conntrack_l3proto nf_conntrack_l3proto_ipv6 __read_mostly = {
374         .l3proto                = PF_INET6,
375         .name                   = "ipv6",
376         .pkt_to_tuple           = ipv6_pkt_to_tuple,
377         .invert_tuple           = ipv6_invert_tuple,
378         .print_tuple            = ipv6_print_tuple,
379         .print_conntrack        = ipv6_print_conntrack,
380         .get_l4proto            = ipv6_get_l4proto,
381 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
382         .tuple_to_nlattr        = ipv6_tuple_to_nlattr,
383         .nlattr_to_tuple        = ipv6_nlattr_to_tuple,
384         .nla_policy             = ipv6_nla_policy,
385 #endif
386 #ifdef CONFIG_SYSCTL
387         .ctl_table_path         = nf_net_netfilter_sysctl_path,
388         .ctl_table              = nf_ct_ipv6_sysctl_table,
389 #endif
390         .me                     = THIS_MODULE,
391 };
392
393 MODULE_ALIAS("nf_conntrack-" __stringify(AF_INET6));
394 MODULE_LICENSE("GPL");
395 MODULE_AUTHOR("Yasuyuki KOZAKAI @USAGI <yasuyuki.kozakai@toshiba.co.jp>");
396
397 static int __init nf_conntrack_l3proto_ipv6_init(void)
398 {
399         int ret = 0;
400
401         need_conntrack();
402
403         ret = nf_ct_frag6_init();
404         if (ret < 0) {
405                 printk("nf_conntrack_ipv6: can't initialize frag6.\n");
406                 return ret;
407         }
408         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_tcp6);
409         if (ret < 0) {
410                 printk("nf_conntrack_ipv6: can't register tcp.\n");
411                 goto cleanup_frag6;
412         }
413
414         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_udp6);
415         if (ret < 0) {
416                 printk("nf_conntrack_ipv6: can't register udp.\n");
417                 goto cleanup_tcp;
418         }
419
420         ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_icmpv6);
421         if (ret < 0) {
422                 printk("nf_conntrack_ipv6: can't register icmpv6.\n");
423                 goto cleanup_udp;
424         }
425
426         ret = nf_conntrack_l3proto_register(&nf_conntrack_l3proto_ipv6);
427         if (ret < 0) {
428                 printk("nf_conntrack_ipv6: can't register ipv6\n");
429                 goto cleanup_icmpv6;
430         }
431
432         ret = nf_register_hooks(ipv6_conntrack_ops,
433                                 ARRAY_SIZE(ipv6_conntrack_ops));
434         if (ret < 0) {
435                 printk("nf_conntrack_ipv6: can't register pre-routing defrag "
436                        "hook.\n");
437                 goto cleanup_ipv6;
438         }
439         return ret;
440
441  cleanup_ipv6:
442         nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv6);
443  cleanup_icmpv6:
444         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmpv6);
445  cleanup_udp:
446         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp6);
447  cleanup_tcp:
448         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp6);
449  cleanup_frag6:
450         nf_ct_frag6_cleanup();
451         return ret;
452 }
453
454 static void __exit nf_conntrack_l3proto_ipv6_fini(void)
455 {
456         synchronize_net();
457         nf_unregister_hooks(ipv6_conntrack_ops, ARRAY_SIZE(ipv6_conntrack_ops));
458         nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv6);
459         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmpv6);
460         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp6);
461         nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp6);
462         nf_ct_frag6_cleanup();
463 }
464
465 module_init(nf_conntrack_l3proto_ipv6_init);
466 module_exit(nf_conntrack_l3proto_ipv6_fini);